1.3. Cluster Infrastructure

Cluster management manages cluster quorum and cluster membership. CMAN (an abbreviation for cluster manager) performs cluster management in Red Hat Cluster Suite for Red Hat Enterprise Linux 5. CMAN is a distributed cluster manager and runs in each cluster node; cluster management is distributed across all nodes in the cluster (refer to 图 1.2 “CMAN/DLM Overview”).

CMAN keeps track of cluster quorum by monitoring the count of cluster nodes. If more than half the nodes are active, the cluster has quorum. If half the nodes (or fewer) are active, the cluster does not have quorum, and all cluster activity is stopped. Cluster quorum prevents the occurrence of a "split-brain" condition — a condition where two instances of the same cluster are running. A split-brain condition would allow each cluster instance to access cluster resources without knowledge of the other cluster instance, resulting in corrupted cluster integrity.

Quorum 由​​​​​​​群​​​​​​​集​​​​​​​节​​​​​​​点​​​​​​​间​​​​​​​经​​​​​​​由​​​​​​​以​​​​​​​太​​​​​​​网​​​​​​​的​​​​​​​消​​​​​​​息​​​​​​​通​​​​​​​信​​​​​​​决​​​​​​​定​​​​​​​的​​​​​​​。​​​​​​​Quorum 也​​​​​​​可​​​​​​​以​​​​​​​由​​​​​​​通​​​​​​​过​​​​​​​以​​​​​​​太​​​​​​​网​​​​​​​和​​​​​​​ quorum 磁​​​​​​​盘​​​​​​​的​​​​​​​消​​​​​​​息​​​​​​​的​​​​​​​组​​​​​​​合​​​​​​​来​​​​​​​决​​​​​​​定​​​​​​​。​​​​​​​对​​​​​​​于​​​​​​​经​​​​​​​由​​​​​​​以​​​​​​​太​​​​​​​网​​​​​​​的​​​​​​​ quorum，quorum 由​​​​​​​节​​​​​​​点​​​​​​​ vote 的​​​​​​​ 50% 加​​​​​​​ 1 组​​​​​​​成​​​​​​​的​​​​​​​。​​​​​​​而​​​​​​​对​​​​​​​于​​​​​​​经​​​​​​​由​​​​​​​ quorum 磁​​​​​​​盘​​​​​​​的​​​​​​​ quorum，quorum 由​​​​​​​用​​​​​​​户​​​​​​​指​​​​​​​定​​​​​​​的​​​​​​​条​​​​​​​件​​​​​​​组​​​​​​​成​​​​​​​。​​​​​​​

CMAN 通​​​​​​​过​​​​​​​监​​​​​​​控​​​​​​​其​​​​​​​他​​​​​​​群​​​​​​​集​​​​​​​节​​​​​​​点​​​​​​​的​​​​​​​信​​​​​​​息​​​​​​​来​​​​​​​跟​​​​​​​踪​​​​​​​成​​​​​​​员​​​​​​​资​​​​​​​格​​​​​​​。​​​​​​​当​​​​​​​群​​​​​​​集​​​​​​​的​​​​​​​成​​​​​​​员​​​​​​​资​​​​​​​格​​​​​​​发​​​​​​​生​​​​​​​变​​​​​​​化​​​​​​​时​​​​​​​，群​​​​​​​集​​​​​​​管​​​​​​​理​​​​​​​器​​​​​​​通​​​​​​​知​​​​​​​其​​​​​​​他​​​​​​​基​​​​​​​础​​​​​​​结​​​​​​​构​​​​​​​组​​​​​​​件​​​​​​​，它​​​​​​​们​​​​​​​再​​​​​​​采​​​​​​​取​​​​​​​合​​​​​​​适​​​​​​​的​​​​​​​行​​​​​​​动​​​​​​​。​​​​​​​例​​​​​​​如​​​​​​​，如​​​​​​​果​​​​​​​节​​​​​​​点​​​​​​​ A 加​​​​​​​入​​​​​​​了​​​​​​​群​​​​​​​集​​​​​​​并​​​​​​​挂​​​​​​​载​​​​​​​了​​​​​​​节​​​​​​​点​​​​​​​ B 和​​​​​​​ C已​​​​​​​经​​​​​​​挂​​​​​​​载​​​​​​​的​​​​​​​ GFS 文​​​​​​​件​​​​​​​系​​​​​​​统​​​​​​​，那​​​​​​​么​​​​​​​节​​​​​​​点​​​​​​​ A 就​​​​​​​需​​​​​​​要​​​​​​​额​​​​​​​外​​​​​​​的​​​​​​​ journal 和​​​​​​​锁​​​​​​​的​​​​​​​管​​​​​​​理​​​​​​​来​​​​​​​使​​​​​​​用​​​​​​​这​​​​​​​个​​​​​​​ GFS 文​​​​​​​件​​​​​​​系​​​​​​​统​​​​​​​。​​​​​​​如​​​​​​​果​​​​​​​群​​​​​​​集​​​​​​​节​​​​​​​点​​​​​​​在​​​​​​​规​​​​​​​定​​​​​​​的​​​​​​​时​​​​​​​间​​​​​​​内​​​​​​​没​​​​​​​有​​​​​​​传​​​​​​​送​​​​​​​消​​​​​​​息​​​​​​​，群​​​​​​​集​​​​​​​管​​​​​​​理​​​​​​​器​​​​​​​将​​​​​​​将​​​​​​​这​​​​​​​个​​​​​​​节​​​​​​​点​​​​​​​从​​​​​​​群​​​​​​​集​​​​​​​里​​​​​​​删​​​​​​​除​​​​​​​，并​​​​​​​通​​​​​​​知​​​​​​​其​​​​​​​他​​​​​​​基​​​​​​​础​​​​​​​结​​​​​​​构​​​​​​​组​​​​​​​件​​​​​​​这​​​​​​​个​​​​​​​节​​​​​​​点​​​​​​​已​​​​​​​经​​​​​​​不​​​​​​​再​​​​​​​是​​​​​​​成​​​​​​​员​​​​​​​了​​​​​​​。​​​​​​​其​​​​​​​他​​​​​​​群​​​​​​​集​​​​​​​基​​​​​​​础​​​​​​​结​​​​​​​构​​​​​​​组​​​​​​​件​​​​​​​在​​​​​​​获​​​​​​​得​​​​​​​这​​​​​​​个​​​​​​​通​​​​​​​知​​​​​​​后​​​​​​​将​​​​​​​决​​​​​​​定​​​​​​​采​​​​​​​取​​​​​​​什​​​​​​​么​​​​​​​样​​​​​​​的​​​​​​​行​​​​​​​动​​​​​​​。​​​​​​​例​​​​​​​如​​​​​​​，Fencing 将​​​​​​​保​​​​​​​护​​​​​​​不​​​​​​​再​​​​​​​是​​​​​​​成​​​​​​​员​​​​​​​的​​​​​​​节​​​​​​​点​​​​​​​。​​​​​​​